Improving gait classification in horses by using inertial measurement unit (IMU) generated data and machine learning.

For centuries humans have been fascinated by the natural beauty of horses in motion and their different gaits. Gait classification (GC) is commonly performed through visual assessment and reliable, automated methods for real-time objective GC in horses are warranted. In this study, we used a full body network of wireless, high sampling-rate sensors combined with machine learning to fully automatically classify gait. Using data from 120 horses of four different domestic breeds, equipped with seven motion sensors, we included 7576 strides from eight different gaits. GC was trained using several machine-learning approaches, both from feature-extracted data and from raw sensor data. Our best GC model achieved 97% accuracy. Our technique facilitated accurate, GC that enables in-depth biomechanical studies and allows for highly accurate phenotyping of gait for genetic research and breeding. Our approach lends itself for potential use in other quadrupedal species without the need for developing gait/animal specific algorithms.

The effect of rider weight and additional weight in Icelandic horses in tölt: part I. Physiological responses.

This study examined the effect of increasing BW ratio (BWR) between rider and horse, in the BWR range common for Icelandic horses (20% to 35%), on heart rate (HR), plasma lactate concentration (Lac), BWR at Lac 4 mmol/l (W4), breathing frequency (BF), rectal temperature (RT) and hematocrit (Hct) in Icelandic horses. In total, eight experienced school-horses were used in an incremental exercise test performed outdoors on an oval riding track and one rider rode all horses. The exercise test consisted of five phases (each 642 m) in tölt, a four-beat symmetrical gait, at a speed of 5.4±0.1 m/s (mean±SD), where BWR between rider (including saddle) and horse started at 20% (BWR20), was increased to 25% (BWR25), 30% (BWR30), and 35% (BWR35) and finally decreased to 20% (BWR20b). Between phases, the horses were stopped (~5.5 min) to add lead weights to specially adjusted saddle bags and a vest on the rider. Heart rate was measured during warm-up, the exercise test and after 5, 15 and 30 min of recovery and blood samples were taken and BF recorded at rest, and at end of each of these aforementioned occasions. Rectal temperature was measured at rest, at end of the exercise test and after a 30-min recovery period. Body size and body condition score (BCS) were registered and a clinical examination performed on the day before the test and for 2 days after. Heart rate and BF increased linearly (P0.05), but negative correlations (P<0.05) existed between body size measurements and Hct. While HR, Hct and BF recovered to values at rest within 30 min, Lac and RT did not. All horses had no clinical remarks on palpation and at walk 1 and 2 days after the test. In conclusion, increasing BWR from 20% to 35% resulted in increased HR, Lac, RT and BF responses in the test group of experienced adult Icelandic riding horses. The horses mainly worked aerobically until BWR reached 22.7%, but considerable individual differences (17.0% to 27.5%) existed that were not linked to horse size, but to back BCS.

The effect of rider weight and additional weight in Icelandic horses in tölt: part II. Stride parameters responses.

This study investigated the effects of rider weight in the BW ratio (BWR) range common for Icelandic horses (20% to 35%), on stride parameters in tölt in Icelandic horses. The kinematics of eight experienced Icelandic school horses were measured during an incremental exercise test using a high-speed camera (300 frames/s). Each horse performed five phases (642 m each) in tölt at a BWR between rider (including saddle) and horse starting at 20% (BWR20) and increasing to 25% (BWR25), 30% (BWR30), 35% (BWR35) and finally 20% (BWR20b) was repeated. One professional rider rode all horses and weight (lead) was added to saddle and rider as needed. For each phase, eight strides at speed of 5.5 m/s were analyzed for stride duration, stride frequency, stride length, duty factor (DF), lateral advanced placement, lateral advanced liftoff, unipedal support (UPS), bipedal support (BPS) and height of front leg action. Stride length became shorter (Y=2.73-0.004x; P0.05). In conclusion, increased BWR decreased stride length and increased DF proportionally to the same extent in all limbs, whereas BPS increased at the expense of decreased UPS. These changes can be expected to decrease tölt quality when subjectively evaluated according to the breeding goals for the Icelandic horse. However, beat, symmetry and height of front leg lifting were not affected by BWR.

A comparison of the physiological response to tölt and trot in the Icelandic horse.

This study compared the effect of ridden tölt and trot at 3 speeds on physiological responses in trained adult (15.3 ± 1.6 yr) Icelandic horses. The experiment had a crossover design with 8 horses, 2 treatments (incremental exercise test in tölt and trot), and 2 riders. Each horse performed 2 tests per day (1 gait with 2 riders, minimum 4.5 h between) on 2 separate days, with 1 d of rest in between. The exercise test consisted of three 642-m phases at 3.0 m/s (Speed), 4.0 m/s (Speed), and 5.0 m/s (Speed) and was performed outdoors on a 300-m oval gravel riding track in northern Iceland in May 2012. Heart rate (HR) was measured during warm-up, the exercise test, and after 5, 15, and 30 min of recovery. Blood samples were taken at rest, after warm-up, after each phase of the exercise test, and after 5, 15, and 30 min of recovery. Respiratory rate was counted for at least 15 s at rest, at the end of the exercise test, and at the end of the 30-min recovery, and rectal temperature was measured on these occasions. There were no differences in HR between tölt and trot at any time point ( > 0.05). At Speed, hematocrit and plasma lactate concentration were greater ( < 0.05) in tölt (40% ± 1%, 1.1 ± 0.06 mmol/L) than in trot (39% ± 1%; 0.9 ± 0.06 mmol/L). There was a prolonged recovery of hematocrit and respiratory rate, a slower decrease in rectal temperature, and a tendency of a prolonged recovery of plasma lactate concentration ( = 0.0675) after tölt. In conclusion, there were only minor differences in physiological responses to tölt and trot in this selected group of experienced adult Icelandic horses and the biological and practical significance of the slightly elevated physiological responses to tölt and the slower recovery remains to be determined.

Physiological response to a breed evaluation field test in Icelandic horses.

This study examined the response in terms of heart rate (HR), respiratory rate (RR), haematocrit (Htc), rectal temperature (RT), and some plasma variables in Icelandic horses of different sexes and ages performing the riding assessment in a breed evaluation field test (BEFT). The study was conducted in Iceland on 266 horses (180 mares and 86 stallions, divided into four age groups; 4, 5, 6 and ≥7 years old). RT and RR were recorded and blood samples were taken before the warm-up and after the riding assessment. Horse HR, velocity and distance were recorded during the warm-up, the riding assessment and a 5-min recovery period. The distance covered in the BEFT was 2.9 ± 0.4 km (range: 1.8 to 3.8 km, n=248), the duration was 9:37 ± 1:22 min:s (range: 5:07 to 15:32 min:s, n=260) and the average speed was 17.8 ± 1.4 km/h (range: 13.2 to 21.3 km/h, n=248). Average HR was 184 ± 13 b.p.m. (range: 138 to 210 b.p.m., n=102) and peak HR 224 ± 9 b.p.m. (range: 195 to 238 b.p.m., n=102), and 36% of the BEFT was performed at HR ≥200 b.p.m. Post-exercise plasma lactate concentration (Lac) was 18.0 ± 6.5 mmol/l (range: 2.1 to 34.4 mmol/l, n=266), and there was an increase in total plasma protein, plasma creatine kinase and aspartate amino transferase concentration, as well as RR, RT and Htc. Stallions covered a longer total distance (in the warm-up and BEFT) (P<0.05), at a faster speed during BEFT (P<0.001) than mares and had higher Htc and lower HR and post-exercise Lac values. There were few effects of age, but the 4- and 5-year-old horses had lower Htc than older horses and 4-year-old horses had higher post-exercise RR than older horses, although they were ridden for a shorter distance, shorter duration and at lower peak velocity (P<0.1). The results showed that the riding assessment in the BEFT is a high-intensity exercise. The results also showed that aerobic fitness was higher in stallions and that age had a limited effect on the physiological response. It is suggested that these results should be used as a guide for the development of training programmes and fitness tests in Icelandic horses that would improve both performance and welfare of the horse.

A survey of the fertility of Icelandic stallions.

Very limited information is available on the breeding performance of Icelandic stallions, let alone the effect that management practices may have had on such performance. As an extensively kept, largely genetically isolated breed of horse it provides a good model for the study of factors that affect reproductive performance without the additional complication of selective breeding, infectious infertility and breed effect. A survey was conducted using 27 Icelandic stallions covering 1590 mares within the normal Icelandic breeding system (May to September). During the season, stallions cover mares within three periods of time, each period being of a similar length (average 35.5 days). During period 1, mares are covered in hand and at pasture. During periods 2 and 3, all mares are covered at pasture. The overall fertility rate for Icelandic stallions was calculated. The effect of a range of variables on fertility was investigated statistically using a number of models in an attempt to minimise the effect of confounding factors. An overall adjusted fertility rate for Icelandic stallions of 67.7% was obtained. The following factors were shown to have a significant effect on fertility: age of mare (P<0.001), training level of stallion (P<0.05) and method of breeding (P<0.05). For some individual stallions reproductive status of the mare also had a significant (P<0.001) effect. Many of these factors have been observed to effect FR in other more intensively managed equine populations. However, the less dramatic detrimental effect of age and the lack of a significant effect of mare reproductive status in most stallions suggests that infertility problems are less evident in Icelandic mares, possibly due to less emphasis on selection for athletic performance and the accepted culling of subfertile stock.